Abstract:  In the insurance industry,customer classification is the basis for differentiated marketing.Due to the knowledge redundancy,the normally adopted traditional clustering method for customer classification suffers from low clustering quality.In order to achieve an effective classification,this paper proposes a clustering model based on attribute reduction and selforganizing feature map (SOM) neural network.By using attribute reduction rule,it can distinguish the redundant knowledge and effectively discover key attributes.Key attributes are used as inputs for learning by SOM neural network so as to obtain better clustering quality.The effectiveness of proposed method is verified through a case study from the H Insurance company.  

Key words: attribute reduction, clustering analysis, self-organizing feature map (SOM) neural network, customer classification